Verbal fluency consists of semantic and phonemic fluency and is often used to detect verbal ability and executive control (Shao et al., 2014). While research has found general verbal fluency impairments in chronic alcohol use, few studies have examined semantic and phonemic fluency separately (Stavro et al., 2012; Stephan et al., 2017). This meta-analytical study examines the performance of abstinent alcohol-dependent individuals on semantic fluency (categories) and phonemic fluency (letters).

As part of a larger study, two researchers independently searched eight databases, extracted required data, and calculated effect sizes on neuropsychological data in alcohol dependent (AD) individuals. Inclusion criteria for articles were: (a) comparison of abstinent alcohol-dependent patients to healthy controls, (b) matched control group on age, education, or IQ, and (c) standardized neuropsychological testing. Exclusion criteria included: (a) diagnosis of Axis I disorders (other than alcohol dependence), (b) comorbidity with other disorders that impact neuropsychological functioning, or (c) not published or translated into English. A total of 31 articles (AD n=1,080 and HC n=1,090) was analyzed in this study.

Semantic fluency evidenced a statistically significant and medium effect size estimate (g = 0.632, p < 0.001). The heterogeneity for semantic fluency was statistically significant (Q=152.468, df=20, p=0.000). Phonemic fluency evidenced a statistically significant and medium effect size estimate (g = 0.572, p < 0.001). The heterogeneity for phonemic fluency was also statistically significant (Q=236.697, df=24, p=0.000).

Deficits in semantic and phonemic fluency are both associated with alcohol dependence. Although some previous research has reported more frontal lobe impact of alcohol, which would be expected to impact phonemic more readily than semantic fluency, this is not evident in the current data. There are many possible reasons for this failure to observe this dissociation meta-analytically. Some potential reasons include the possibility that alcohol affects multiple regions of the brain, that both these measures are affected by alcohol but miss the subtlety associated with frontal damage, or the likelihood that when studies are aggregated in meta-analysis the heterogeneity results in a regression to the mean effect size. These and other reasons are not mutually exclusive and future research should attempt to examine these and other hypotheses.

Health literacy is the degree to which an individual is able to attain, process, and understand information, skills, and services required to make informed decisions. Limited health literacy is a risk factor for problems understanding health information and adhering to medical instructions, underuse of preventive services, increased hospitalizations and associated medical costs, and higher mortality rates. Recognizing individuals with reduced health literacy can be difficult given demographic information such as age or years of education do not reliably reflect an individual’s health literacy level. Cross-sectional studies have identified limited health literacy as associated with lower scores on cognitive tests measuring memory, executive function (EF), and processing speed, independent from the influence of demographic variables (e.g., age, race, education). This study assessed the association of objective measures of executive functioning and health literacy when controlling for premorbid estimated intellectual functioning and relevant demographic variables.

A sample of 44 adult patients (20 Male; 24 Female) referred for neuropsychological evaluation for memory complaints who were administered the Test of Premorbid Functioning (TOPF), and multiple measures of EF including the Trail Making Test - Part B (TMT-B), Stroop Color and Word Test (SCWT), and Letter (FAS) and Semantic (Animals) Fluency as part of part of a larger standardized battery. Patients were also administered the Short Assessment of Health Literacy-English (SAHL-E). All included patients had <2 performance validity test failures. The sample was 50% non-Hispanic Black, 31.8% non-Hispanic White, 15.9% Hispanic, 2.3% Asian/Pacific Islander, and 54.5% female. Diagnostically, 50.9% of the sample were cognitively normal, 36.4% had mild cognitive impairment, and 15.9% had dementia. Two multiple regressions were conducted to evaluate (1) the predictive power of EF on the SAHL-E, and (2) the moderating impact of estimated premorbid IQ and demographics via the TOPF on the relationship between EF and SAHL-E.

The first regression was not significant (p=.168) with FAS as the only independent predictor of SAHL-E performance (ß=.387, p<.05). The overall model was significant with the addition of the TOPF (p<.001). FAS accounted for 29% (ß=.336, p<.05) of the variance in SAHL-E when controlling for TOPF and other measures of EF.

These results indicate that novel generativity is a significant predictor of health literacy beyond the influence of estimated premorbid intelligence and demographic factors. Importantly, these findings suggest that broadly speaking EF abilities have minimal impact on health literacy, although reduced verbal generativity to letter cues is associated with reduced health literacy. Identification of at-risk populations such as individuals with limited health literacy is clinically important and can make way for early intervention. Health information targeted at this at-risk population should go beyond vocabulary and more specifically reduce the burden on verbal fluency.

When asked to imagine future events, individuals with PTSD provide narratives with limited event-specific details, suggesting an impairment in event elaboration. Here we examined whether future thinking in PTSD is also associated with an impairment in the initial stage of event construction, by using a future-event fluency task that makes no demands on event elaboration (MacLeod, A. K., & Salaminiou, E. (2001). Reduced positive future-thinking in depression: Cognitive and affective factors. Cognition & Emotion, 15(1), 99–107).

Thirty-five veterans (6 female, 29 male; aged 27–51), assigned on the basis of structured diagnostic interviews to PTSD-only (n = 15), PTSD + depression (n = 9), or psychopathology-free control groups (n = 11), were asked to generate, in one minute, as many events as possible that they expected to happen in the future, across four conditions that varied in valence (positive, negative) and temporal framework (1 month, 10 years). Two independent raters classified each event generated as being specific (i.e., a unique, time-limited event), generic (i.e., ongoing or recurring events), or a repetition.

Results of linear mixed modeling carried out on the number of specific events generated showed that diagnostic group and event valence contributed significantly to the overall model fit. All participants generated more positive than negative events (β = 1.014, SE = 0.330, t(105) = 3.07 p = 0.003), and both PTSD groups generated fewer specific events than controls (PTSD-only (β = -2.203, SE = 0.744, t(35) = -2.96, p = 0.005); PTSD + depression (β = -1.859, SE = 0.842, t(35) = -2.21, p = 0.034). Adding the interaction between group and valence did not improve the model fit, suggesting that the PTSD groups were not differentially impaired in the generation of positive and negative events. When including scores on an emotionally neutral phonemic fluency task (FAS) as a covariate to account for verbal fluency, the PTSD-only group still generated significantly fewer events than the controls (β = -1.667, SE = 0.733, t(34) = -2.27, p = 0.030). After adjusting for FAS, the group effect was marginal for the PTSD + depression group (β = -1.600, SE = 0.801, t(34) = -2.00, p = 0.054).

These results suggest that the impairment in future thinking in PTSD concerns not only the elaboration of future events but also the processes involved in initial event specification, such as those involved in the search and selection of a specific event. Moreover, these findings highlight a distinction between the future thinking abnormalities in PTSD, characterized by reduced generation of both positive and negative future events, compared to depression, which has been associated with reduced generation of positive future events only (MacLeod & Salaminiou, 2001).

Late-life depression is prevalent among older adults and the presence of depressive symptoms has been shown to be associated cross-sectionally with worse verbal fluency performance. There is limited and mixed evidence as to whether depressive symptoms impact change in verbal fluency performance over time, and whether gender impacts this relationship.

Participants were community-dwelling older adults who were dementia-free at baseline (N = 522; M age = 75.96, SD + 6.46 years). Baseline depressive symptoms were measured using the Geriatric Depression Scale. Category fluency and letter fluency performance, using the Controlled Oral Word Association Test (COWAT), were examined annually. Linear mixed effects models stratified by gender examined whether associations between baseline depressive symptoms and changes in fluency performance over five years were different in female (n = 289) as compared to male (n = 233) participants. Sensitivity analyses excluding participants with prevalent or incident mild cognitive impairment (MCI) (n = 141), excluding participants with incident dementia (n = 28), and excluding participants with prevalent or incident MCI or incident dementia (n = 169) were run. All analyses were adjusted for age, years of education, estimated premorbid functioning, and health comorbidities.

Depression was minimal across participants (m = 4.72, SD + 3.96). A subset of participants (n = 44) reported “possible depression,” namely levels suggestive of subclinical depression, according to clinical cutoffs. The “possible depression” group included 31 females (10.73% of females) and 13 males (5.58% of males), and the “no depression” group included 258 females (89.27% of females) and 220 males (94.42% of males). Baseline levels of depressive symptoms suggestive of subclinical depression were associated with worse decline in category fluency performance during longitudinal follow-up in females (estimate = -0.16, p = .002) but not males (estimate = -0.03, p = .658). Results remained the same when excluding prevalent and incident MCI cases (estimate = -0.19, p = .005), excluding incident dementia cases (estimate = -0.12, p = .017), and excluding prevalent and incident MCI and incident dementia cases (estimate = -0.20, p = .004). Letter fluency performance did not decline over time and was not influenced by levels of depressive symptoms in females (estimate = -0.03, p = .502) or males (estimate = 0.05, p = .452).

Baseline presence of depressive symptoms suggestive of subclinical depression was associated with worse decline in category fluency performance during longitudinal follow-up in female but not male participants. Letter fluency performance did not decline and was not impacted by levels of depressive symptoms. Results remained significant when accounting for covariates and potential confounders. The present study elucidated the combined influence of gender and depressive symptoms on change in fluency performance in older adults and can aid in identifying individuals who may be at a greater risk of cognitive decline.

Word finding difficulty is a prevalent cognitive symptom in multiple sclerosis (MS). Word finding relies on retrieving concepts and word forms from the long-term store. Neuropsychological assessment of word finding difficulty in persons with MS (pwMS) is typically characterized by semantic errors and decreased speed in naming tests, along with decreased semantic verbal fluency scores. Despite this, there is significant heterogeneity in the detection of verbal fluency deficits across studies in the MS literature. This may be partially due to disease-related heterogeneity and/or low sensitivity of commonly used scoring approaches. We investigate the latter in the present study. Semantic network analysis, derived from graph theory, provides a fine-grained approach to understanding semantic retrieval by utilizing information about the co-occurrence of words produced on semantic verbal fluency tasks. Analysis results in a graphical quantification of the conceptual-lexical store. A preliminary study found that semantic networks from Spanish-speaking pwMS had fewer associative connections and more central connective pathways, which if affected, may lead parts of the network to become inaccessible for retrieval. However, their investigation was limited in the generalizability of their findings, as they excluded pwMS who have cognitive impairment (CI), which represents a significant proportion of pwMS. We sought to investigate network differences in an English-speaking MS sample, without exclusion based on CI, using widely-used metrics of micro-, meso-, and macroscopic structure. We hypothesize the MS network will be less efficiently organized, thus characterized by higher average shortest path length (ASPL), lower clustering coefficient (CC) and lower modularity (Q).

53 persons with MS and 44 neurologically healthy controls (HC) were recruited as a part of an ongoing study (NMSS RG-1907-34364 & RG-1901-33304). As a part of a larger battery, participants were administered the semantic verbal fluency subtest of the Controlled Oral Word Association Test. Responses were analyzed using a network-analysis R suite.

The MS and HC networks were characterized by having similar average shortest path lengths (ASPL MS =2.466, ASPL HC=2.463, F(1,1997)=0.281, p=0.596), indicating they require similar numbers of edges to be traversed to reach other nodes in the network. This suggests similar efficiency of information transfer. Clustering coefficient was not significantly different between the MS and HC networks (CC MS = 0.742, CC HC =0.742, F(1,1997)=0.10, p=0.919), suggesting similar local interconnectivity. The MS network had significantly lower modularity compared to the HC network (Q MS =0.497, Q HC = 0.502, F(1,1997)=16.678, p<0.001). This means that sub-communities of the network were less segregated into densely connected sub-graphs.

Contrary to expectation, ASPL and CC were not significantly different between groups. The absence of finding lower CC was consistent with prior findings. Consistent with our hypothesis, the MS network had lower modularity. This may suggest that pwMS were unable to use categorical clustering to aid in retrieval from the lexicon. Specifically, low modularity coupled with similar CC may suggest the structure of the MS lexicon is characterized by intact clustering on a microscopic scale but less strong organization into distinct clusters on a larger scale.

To determine if the degree of split between phonemic verbal fluency and semantic verbal fluency at initial visit is predictive of decline in memory performance between initial evaluation and follow-up.

Data from a retrospective multidisciplinary memory clinic database at Spectrum Health was utilized. We examined data from 90 participants who had both an initial and follow-up evaluation completed (initial age = 77.1±4.7 years, follow-up age = 78.4±4.7 years, education = 13.9±3.1 years, race = 91% White, 7% Black, & 2% Hispanic, sex = 61% female, time between evaluations = 15.2±9.9 months). Patients who returned for follow-up did not meet criteria for dementia at time one. Split between phonemic and semantic fluency, termed the semantic-phonological delta (SPD) was measured at the initial evaluation by subtracting the Controlled Oral Word Association Test (COWAT; FAS) T-score from the Animal Naming Test (ANT) T-score. Change in memory score was defined in two ways: 1) subtracting the follow-up evaluation Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) List Recognition Score (RBANS percentiles were converted to a scaled score) from the initial evaluation List Recognition Score (List Recognition Delta), and 2) computing the difference in the RBANS Delayed Memory Index Standard Score between the initial evaluation and the follow-up evaluation (RBANS Memory Delta).

Average semantic fluency T scores were (M = 40.3, SD = 12.3) and phonemic fluency T scores were (M = 42.7, SD = 10.3) at initial evaluation. Bivariate correlations were used to determine the relationship between the clinical variables. SPD was significantly correlated with List Recognition Delta, r(88) = .23, p = .026, with greater discrepancies in verbal fluency scores associated with higher level of decline in List Recognition at follow-up. By comparison, Semantic Fluency performance itself at initial evaluation was not significantly correlated with List Recognition Delta, r(88) = .17, p = .097. The correlation between SPD and the RBANS Memory Delta was also not significant, r(88) = .14, p = .166. At follow-up evaluation, 39% of the sample received a diagnosis of Alzheimer’s disease. Of those diagnosed with Alzheimer’s disease, 66% had a negative SPD split at time one, performing worse on semantic fluency compared to phonemic fluency.

SPD is a better predictor of decline in RBANS List Recognition performance between evaluations than semantic fluency alone, with a larger negative SPD score (worse semantic fluency performance compared to phonemic fluency) at initial evaluation predicting decline in List Recognition performance at follow-up evaluation. SPD at initial evaluation was not significantly correlated with change in RBANS Delayed Memory Index score between evaluations. This may be because there are some patients who are similarly impaired in both semantic and phonemic verbal fluency at initial evaluation who later demonstrate progressive decline in memory retrieval due to hippocampal-sparing etiologies (e.g., vascular dementia). Overall, these findings are consistent with previous work suggesting that declines in the semantic memory system precede declines in episodic memory retention in conditions such as Alzheimer’s disease.

As neuropsychologists aim to collect valid data, maximize the utility of assessments, make effective use of time, and best serve patient populations, measurement of performance validity is considered a critical issue for the field. As effort may vary across an evaluation, including performance validity tests (PVTs) throughout the assessment is important. Incorporating embedded PVTs in addition to free standing PVTs can be particularly useful in this regard. COWAT and animal naming are commonly administered verbal fluency measures. While there have been past investigations into their potential for detecting invalid performance, they are limited, and more research is needed. Perhaps most promising, Sugarman and Axelrod (2015) described a logistic regression derived formula utilizing the combined raw scores of COWAT and animal naming. The current study aimed to investigate the use of embedded PVTs within COWAT and animal naming to provide further support for the use of embedded PVTs in these measures.

All subjects were from a mixed clinical sample comprising military veterans from two VA Medical Centers in the northeast U.S., who were referred for neuropsychological evaluation. Subjects deemed credible had zero PVT failures. Subjects were considered non-credible performers if they failed at least two out of a possible eight PVTs administered. Subjects who failed one PVT were excluded from the study (n = 53). The final sample consisted of 116 individuals with credible performance (Mean Age = 35.5, SD = 8.8; Mean Edu = 13.6, SD = 2; Mean Est. IQ = 106, SD = 7.9) and 94 individuals with psychometrically determined non-credible performance (Mean Age = 38.5, SD = 9.4; Mean Edu = 113, SD = 2.1; Mean Est. IQ = 101, SD = 8.7). Performance of COWAT and animals in detecting non-credible performances was evaluated through calculation of classification accuracy statistics and use of the logistic regression formulas reported in Sugarman and Axelrod (2015).

For COWAT, the optimal cutoff was a raw score of <27 (specificity = 89%; sensitivity = 31%), and a T-score of <35 (specificity = 92%; sensitivity = 31%). For animal naming, optimal cutoffs were <16 for raw score (specificity = 92%, sensitivity = 38%) and <37 for T-score (specificity = 91%; sensitivity = 33%). The logistic regression formula based on raw scores for both COWAT and animal naming was inadequately sensitive at the recommended cutoff in this sample, but a coefficient of > .28 was revealed to be optimal (91% specificity; 42% sensitivity). When the formula for T-scores was used, a coefficient of > .38 was optimal (91% specificity; 28% sensitivity).

Results of the current research suggest that PVTs embedded within the commonly administered COWAT and animal naming verbal fluency tests can effectively detect low effort, in concordance with generally accepted standards. A logistic regression formula using raw scores in particular appears to be most effective, consistent with findings reported by Sugarman and Axelrod (2015).

Published results focusing on language assessment in acutely recovered COVID-19 patients have shown communication problems in this group, including significant cognitive-linguistic disruptions in verbal fluency (Cummings, 2022). Extant research also indicates that poorer health-related outcomes, such as reduced physical functioning and quality of life, co-occur with cognitive difficulties post-COVID-19 infection (Mendez et al., 2021; Tabacof et al., 2022). Understanding what factors may worsen the impact of COVID-19 on cognition, and aspects of language function specifically, is necessary to determine who is at greatest risk of adverse outcomes following infection. Our goal was to examine the effect of health-related outcomes on language abilities, specifically verbal fluency, post-COVID-19 infection.

37 adults 19 years and older (M age = 38.78, 67.5% female, 92.5%> high school education) were recruited from British Columbia and Ontario, Canada. Participants provided documentation indicating they had had a COVID-19 infection at least 3 months prior to participation. Participants completed a series of online questionnaires, including the Short Form Health Survey (SF-20), to measure aspects of health-related quality of life. The SF-20 measures dimensions of functioning (physical, social, role) and well-being (mental health, health perception, pain). For each parameter except pain, higher scores indicate better functioning/well-being; for pain higher scores indicate greater pain levels. Participants also completed neuropsychological tests, including measures of verbal fluency, via teleconference. Animals and F-A-S total scores were combined to represent verbal fluency (semantic and phonemic, respectively) performance. To assess the impact of health outcomes on verbal fluency performance, hierarchical regression analyses were conducted. The six SF-20 subscale scores were entered as predictors and verbal fluency score (sum) as the outcome. Age and sex (Male/Female) were controlled for in the model.

Age and sex were not significantly related to verbal fluency scores in our sample. After controlling for these demographics, the overall model including SF-20 subscales did not significantly predict fluency performance (F (8, 28) = 1.04, p = .433). However, Pain scores did individually predict verbal fluency performance (B = 5.60, t = 2.53, p = <.05). Unexpectedly, pain ratings were positively associated with fluency scores, such that each increase in pain rating (e.g., “none” to “mild”) was associated with a fluency score increase of 5.60 points (i.e., 5.6 more words stated across administered tasks).

These preliminary findings suggest that participants’ self-reported pain severity was positively associated with verbal fluency task performance in our sample (i.e., greater pain severity predicting better fluency). These findings are contrary to substantial evidence showing the deleterious effects of pain on cognitive functions in other populations (Khera & Rangasamy, 2021). It is possible that findings may be explained by a potential unknown intervening variable not included in our model. This is the first study to our knowledge to examine associations between experienced pain and verbal fluency performance post-COVID-19 infection. It will be important for future work to not only utilize more robust measures of pain experiences and explore more areas of cognition and language, but also to employ larger samples and examine a broader set of covariates.

Functional magnetic resonance imaging (fMRI) research has generally focused on drawing conclusions from average brain activation patterns. Importantly, the brain is inherently variable; growing literature has found that within-individual blood oxygen level-dependent (BOLD) signal variability may be meaningful, and not just “noise.” For example, recent research has identified increased BOLD signal variability in healthy younger and older adults during more effortful/complex task loads of n-back paradigms, commonly used tasks that involve important elements of executive function (e.g., attention, working memory, planning, inhibition, etc.). Verbal fluency is a complex cognitive domain that also involves similar processes to generate words given certain rules. As a result, the current study builds on existing literature to investigate within-individual BOLD signal variability patterns in peak coordinates of a verbal fluency network during different loads of a letter n-back task. Due to greater executive demands, greater variability was expected during more effortful/complex n-back task loads in regions of a verbal fluency network.

Forty-eight healthy young adults (Mage(SD) = 22.41(4.47), 25 females) from the Atlanta area completed a letter n-back task in an MRI scanner. After standard processing in AFNI, images were corrected for motion and physiological artifacts, which may be confounding sources of variability. Volumes associated with each load of the letter n-back task (0-back, 1-back, 2-back, 3-back, crosshair) were identified. Task runs were normalized and respective run means were subtracted prior to concatenating all runs for each load type. Standard deviations were calculated across this mean-run corrected time series. Ten peak regions of interest (ROIs) were identified from a verbal fluency network generated from 84 peer-reviewed publications for this domain gathered on NeuroSynth. Paired samples t-tests with Benjamini-Hochberg correction for multiple comparisons were conducted to explore differences in variability during n-back task loads.

In several of the verbal fluency network ROIs, within-individual BOLD signal variability was significantly greater for 2-back versus 0-back loads with medium to large effect sizes (p’s < .001 - < .01, Cohen’s d range: .53-.93). Variability was also significantly greater for 3-back versus 0-back loads with small to medium effect sizes (p’s < .001 - < .01, Cohen’s d range: .48-.74). Specific regions that evidenced this pattern included ROIs in the left inferior frontal gyrus, left cingulate, right inferior frontal gyrus, left middle frontal gyrus, and left superior parietal lobule. Only two regions demonstrated increased variability in the 1-back load versus crosshair (left middle frontal gyrus, p < .001, d = .63; left lentiform nucleus, p < .05, d = .42). No regions demonstrated a significant difference in variability in the 0-back load versus crosshair.

This study contributes to growing literature examining within-individual BOLD signal variability in healthy individuals by exploring variability patterns in a verbal fluency network. The observed pattern of results supports the hypothesis and is in line with previous research, demonstrating that greater variability occurs with greater executive task demands. Future research can use an inscanner task of verbal fluency and can extend variability findings during this in-scanner task to out-of-scanner measures of verbal fluency.

There is a pressing need for sensitive, non-invasive indicators of cognitive impairment in those at risk for Alzheimer’s disease (AD). One group at an increased risk for AD is APOEε4 carriers. One study found that cognitively normal APOEε4 carriers are less likely to produce low frequency (i.e., less common) words on semantic fluency tasks relative to non-carriers, but this finding has not yet been replicated. This study aims to replicate these findings within the Wake Forest ADRC clinical core population, and examine whether these findings extend to additional semantic fluency tasks.

This sample includes 221 APOEε4 non-carriers (165 females, 56 males; 190 White, 28 Black/African American, 3 Asian; Mage = 69.55) and 79 APOEε4 carriers (59 females, 20 males; 58 White, 20 Black/African American, 1 Asian; Mage = 65.52) who had been adjudicated as cognitively normal at baseline. Semantic fluency data for both the animal task and vegetable task was scored for total number of items as well as mean lexical frequency (attained via the SUBTLEXus database). Demographic variables and additional cognitive variables (MMSE, MoCA, AMNART scores) were also included from the participants’ baseline visit.

APOEε4 carriers and non-carriers did not differ on years of education, AMNART scores, or gender (ps > 0.05). APOEε4 carriers were slightly younger and included more Black/African American participants (ps < 0.05). Stepwise linear regression was used to determine the variance in total fluency score and mean lexical frequency accounted for by APOEε4 status after including relevant demographic variables (age, sex, race, years of education, and AMNART score). As expected, demographic variables accounted for significant variance in total fluency score (p < 0.0001). Age accounted for significant variance in total fluency score for both the animal task (ß = -0.32, p <0.0001) and the vegetable task (ß = -0.29, p < 0.0001), but interestingly, not the lexical frequency of words produced. After accounting for demographic variables, APOEε4 status did not account for additional variance in lexical frequency for either fluency task (ps > 0.05). Interestingly, APOEε4 status was a significant predictor of total words for the vegetable semantic fluency task only (ß = 0.13, p = 0.01), resulting in a model that accounted for more variance (R2 = 0.25, F(6, 292) = 16.11, p < 0.0001) in total words than demographic variables alone (R2 = 0.23, F(5, 293) = 17.75, p < 0.0001).

Unsurprisingly, we found that age, AMNART, and education were significant predictors of total word fluency. One unexpected finding was that age did not predict the lexical frequency - that is - regardless of age, participants tended to retrieve words of the same lexical frequency, which stands in contrast to the notion that retrieval efficiency of infrequent words declines with age. With regard to APOEε4, we did not replicate existing work demonstrating differences in lexical frequency and semantic fluency tasks for ε4 carriers and non-carriers; possibly due to differences in the demographic characteristics of the sample.